While
seismic processors have long used spectral decomposition, it is
only in recent years that it has been applied directly to aspects
of 3-D seismic data interpretation.
The method
for doing this was first published in "The Leading Edge" in 1999,
in a paper by Greg Partyka and others that illustrated the idea
of using frequency to "tune-in" bed thickness.
Although
spectral decomposition is a relatively new technique, some companies
are experiencing great success in many basins around the world.
(Most of
the best examples are in clastic environments where depositional
stratigraphy is a key driver.)
Companies
using spectral decomposition observe significant detail from these
images at great depth — but have found that interpretation and
integration with well data and models is critical to its success.
As shown
in the channel system of figure 1, spectral
decomposition can extract detailed stratigraphic patterns that help
refine the geologic interpretation of the seismic.
The concept
behind spectral decomposition is that a reflection from a thin bed
has a characteristic expression in the frequency domain that is
indicative of temporal bed thickness.
In other
words, higher frequencies image thinner beds, and lower frequencies
image thicker beds.
This approach
is similar to how remote sensing uses sub-bands of frequencies to
map interference at the earth's surface. Just like remote sensing,
it is very important to dynamically observe the response of the
reservoir to different frequency bands.
The key
is to create a set of data cubes or maps, each corresponding to
a different spectral frequency, which can be viewed through animation
to reveal spatial changes in stratigraphic thickness. Spectral decomposition
reveals details that no single frequency attribute can match.
Based on
well-understood principals, typical amplitude maps are dominated
by the frequency content of seismic data and will best image stratigraphy
with thickness related to the dominant frequencies processed with
the seismic.
This is
illustrated in figure 2a, where we have
a stratigraphic feature that varies in thickness. If the frequency
content is high, thinner stratigraphic features will be "tuned"
in and highlighted by higher amplitude (figure
2b). If the frequency content is lower, thicker stratigraphic
features will stand out (figure 2c).
What is
needed is to see all the different stratigraphic thicknesses in
a meaningful way.
Spectral
decomposition provides this by generating a series of maps or cubes
that observe the response of the reservoir to different frequencies.
These are then animated allowing the interpreter's eye to catch
subtle changes in the reservoir through motion.
There are
other good methods that can analyze tuning, but none are as easy
to create or as routinely used as the method of animation called
the "Tuning Cube."
To use
spectral decomposition, you would interpret a seismic horizon and
create a seismic amplitude map. The amplitude map is critical as
a base to determine if spectral decomposition is adding to your
interpretation.
If you
believe that amplitude is a meaningful indicator for reservoir presence,
then spectral decomposition is a new step in the interpretation
workflow.
The seismic
horizon is then used to transform a window of the data around the
event of interest into the frequency domain and generate a series
of amplitude maps at different frequencies. Thin bed interference
will cause notches in the frequency domain related to the bed's
thickness. This is expressed on the amplitude maps as areas of high
and low amplitude when animating through the different maps.
Subtle
changes in reservoir thickness or internal heterogeneities can be
observed when comparing these images. Very quickly you will get
a feel for areas with active stratigraphic variation that need to
be evaluated in more detail.
Tracking
between these maps and the seismic cross-section is critical to
determine if the features you are seeing are geologically meaningful.
So is combining these images together.
For example,
consider figure 3 , which contains a
stratigraphic feature that appears to have a fan geometry.
At lower
frequencies from the "Tuning Cube," the feeder channel of the "fan"
is highlighted (left image). At higher frequencies, different lobes
of the fan geometry are highlighted (middle image). At the highest
frequencies available in the seismic data, the thinnest areas are
highlighted (right image).
In this
example, there are actually 30 images that need to be animated to
allow the eye to catch all of the detail available. Integration
with well control is critical to determining the accuracy of the
geologic interpretations.
As mentioned,
spectral decomposition is a relatively new technique that already
has helped bring great success in many basins around the world.
As such,
it is poised to become an essential tool for the geologic interpretation
of seismic data.